body weight and risk of colon cancer

American Journal of Epidemiology

ª The Author 2011. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of

Public Health. All rights reserved. For permissions, please e-mail: [email protected].

Vol. 174, No. 10

DOI: 10.1093/aje/kwr247

Advance Access publication:

October 7, 2011

Original Contribution

Body Size and Colorectal Cancer Risk After 16.3 Years of Follow-up: An AnalysisFrom the Netherlands Cohort Study

Laura A. E. Hughes*, Colinda C. J. M. Simons, Piet A. van den Brandt, R. Alexandra Goldbohm,Manon van Engeland, and Matty P. Weijenberg

* Correspondence to Dr. Laura A. E. Hughes, Department of Epidemiology, Faculty of Health Medicine and Life Science, Maastricht

University, P.O. Box 616, 6200 MD, Maastricht, the Netherlands (e-mail: [email protected]).

Initially submitted February 4, 2011; accepted for publication June 22, 2011.

A large body size may differentially influence risk of colorectal cancer (CRC) by anatomic location. The Nether-lands Cohort Study includes 120,852 men and women aged 55–69 years who self-reported weight, height, andtrouser/skirt size at baseline (1986), as well as weight at age 20 years. Derived variables included body mass index(BMI; weight (kg)/height (m)2), BMI at age 20 years, and BMI change. After 16.3 years of follow-up (1986–2002),2,316 CRC cases were available for case-cohort analysis. In men, the highest risk estimates were observed for bodyfat (per 5-unit increase in BMI, hazard ratio (HR)¼ 1.25, 95% confidence interval (CI): 1.05, 1.46; for highest quintileof trouser size vs. lowest, HR ¼ 1.63, 95% CI: 1.17, 2.29 (P-trend ¼ 0.02)) and appeared more closely associatedwith distal colon tumors (for BMI (5-unit increase), HR¼ 1.42, 95%CI: 1.13, 1.79; for highest quintile of trouser size,HR ¼ 2.56, 95% CI: 1.55, 4.24 (P-trend < 0.01)) than with proximal colon or rectal tumors. In women, body fat wasnot associated with CRC risk unless it was considered simultaneously with physical activity; a large trouser/skirt sizeand a low level of physical activity increased risk for all subtypes. Height was associated with risk of CRC, especiallydistal colon tumors (highest quintile vs. lowest: HR ¼ 1.53, 95% CI: 1.03, 2.27; P-trend ¼ 0.05), in women only.

body height; body mass index; cohort studies; colonic neoplasms; rectal neoplasms; waist circumference

Abbreviations: BMI, body mass index; CI, confidence interval; CRC, colorectal cancer; HR, hazard ratio; NLCS, NetherlandsCohort Study.

There is convincing evidence that a large body size in-creases the risk of colorectal cancer (CRC) (1). CRC risk hastraditionally been studied according to anatomic location inthe colon or rectum. However, several anatomic, embryologic,and physiologic differences exist between subanatomic lo-cations in the colorectal tract, and therefore arising tumorsmay have different etiologic pathways (2–5). Thus, it may bemore rational to consider CRC risk at 3 different locations:the proximal colon, distal colon, and rectum. Case-controland prospective cohort data suggest that in both men andwomen, body mass index (BMI) is more strongly associatedwith tumors of the distal colon than with tumors of the prox-imal colon (6–13). Several studies have suggested that waistcircumference, as a proxy measure for abdominal obesity,may be a better indicator of CRC risk than BMI (14, 15) andthat adult attained height as a proxy measure for early-lifenutritional and socioeconomic exposures is also a convincing

risk factor for CRC (1). How other indicators of body size,such as waist circumference and height, might influenceCRC risk at different tumor subsites is less clear.

At present, one can only hypothesize as to how indicatorsof body size may influence subsite-specific pathways of co-lorectal carcinogenesis. Approximately 85% of tumors arethought to develop via the traditional adenoma-carcinomapathway, characterized by mutations in the KRAS oncogeneand the APC tumor suppressor gene, as well as chromosomalinstability (16). Such tumors are most frequently observedin the distal colon (3). The other 15% of tumors are thought toarise via the serrated neoplasia pathway, characterized in partby a high degree of epigenetic instability and microsatelliteinstability (16). Such tumors are most frequently observed inthe proximal colon (17–19). It has been reported that adultBMI and indicators of abdominal obesity are associatedmore strongly with tumors that are not characterized by

1127 Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011

http://aje.oxfordjournals.org/D

ownloaded from

http://aje.oxfordjournals.org/

epigenetic instability (20–22). This suggests that adult bodyfat rather influences CRC risk via chromosomal instabilityand may explain the stronger associations observed betweenbody fat and distal tumors.

Furthermore, timing of exposure may be important fordefining pathways. It has been observed that a high BMI inearly adulthood and a large BMI change from early adulthoodto later adulthood are associated with a higher risk of CRC(13); that a high BMI at age 20 years is associated with agreater risk of having a tumor characterized by epigeneticinstability (21); and that taller persons have a higher risk oftumors characterized by epigenetic instability (Hughes et al.,Maastricht University, unpublished manuscript).

Here, we used data from a prospective study, the Nether-lands Cohort Study (NLCS), to investigate the association be-tween multiple indicators of body size and sex-specific riskof CRC, as well as the associations between these indicatorsand sex-specific CRC risk at specific tumor subsites, includ-ing the proximal colon, distal colon, rectosigmoid junction,and rectum. We considered BMI and trouser/skirt size asadult indicators of body fat, and we also considered adultattained height, BMI at age 20 years, and BMI change whenassessing risk in order to study whether timing of exposurewas associated with differential risk. We hypothesized thatbody fat would be associated more with tumors of the distalcolon, whereas height might be associated more with proximalcolon tumors.

MATERIALS AND METHODS

Study population and design

The NLCS includes 58,279 men and 62,573 womenaged 55–69 years at baseline (1986) who completed a self-administered questionnaire involving 150 food items, as wellas questions on lifestyle and health. Municipal registries fromthroughout the Netherlands were used to constitute an efficientsampling frame (23). The NLCS uses a case-cohort approachfor data processing and analysis; case subjects were derivedfrom the entire cohort, and the number of person-years at riskfor the entire cohort was estimated from a subcohort of5,000 persons who were randomly sampled from the fullcohort at baseline. All subcohort members who reportedprevalent cancer (excluding skin cancer) at baseline were ex-cluded from the analyses, leaving 4,654 participants. Furtherdetails of the NLCS design have been published elsewhere(23–25).

Incident CRC cases were identified by annual record link-age to 9 regional cancer registries and a national pathologydatabase (PALGA) (26). The completeness of cancer follow-up in the NLCS is almost 100% (27). CRCs were classified byanatomic location as proximal colon (International Classifi-cation of Diseases for Oncology, First Edition, codes 153.0,153.1, 153.4, 153.5, and 153.6), distal colon (codes 153.2,153.3, and 153.7), rectosigmoid (code 154.0), or rectal (code154.1) tumors. Persons with incomplete or inconsistent base-line questionnaires were excluded. After additionally exclud-ing persons who were missing data on variables adjusted forin the multivariate analysis, 3,197 subcohort members and2,316 CRC cases remained. Among men, the numbers of

tumors by location were 327 proximal colon, 427 distalcolon, 33 unspecified colon, 125 rectosigmoid junction, and299 rectum; for women, the numbers were 459 proximalcolon, 327 distal colon, 27 unspecified colon, 87 rectosigmoid,and 205 rectum.

Assessment of anthropometric variables, diet, andphysical activity

Height (cm), body weight (kg), and body weight at age20 years (kg) were self-reported on the baseline questionnaire.BMI was calculated as weight (kg)/height (m)2. Participantswere also asked to report their lower-body clothing (trouseror skirt) size from their clothing label (in Dutch sizes).Trouser/skirt size has been shown to be an adequate proxymeasure for waist circumference when predicting cancerrisk in the NLCS (28).

Both occupational physical activity in the longest-/last-heldjob and baseline nonoccupational physical activity were usedto assess risk, depending on sex (Simons et al., MaastrichtUniversity, unpublished manuscript). For analyses of men,occupational physical activity was used. Participants wereasked to report job title(s) and job duration(s) on the base-line questionnaire. Assessment of physical activity at workwas based on the job held for the longest amount of time.Total energy expenditure was based on a rating system devel-oped by Hettinger et al. (29). Men were classified into 3 energyexpenditure groups: <8 kJ/minute, 8–<12 kJ/minute, and�12 kJ/minute. For analyses with women, nonoccupationalphysical activity was used to assess risk, because mostwomen of this generation had not held a job or had workedfor only a short period of time. It was calculated by totalingthe number of minutes per day spent cycling/walking to work,going to shops, and walking the dog and the number of hoursspent per week engaging in gardening/odd jobs, recreationalcycling/walking, and sports/exercise, as reported on thebaseline questionnaire. Women were classified as having alow (<30 minutes/day), intermediate (30–90 minutes/day),or high (>90 minutes/day) level of physical activity.

Statistical analyses

Data were analyzed with Stata, version 10 (StataCorp LP,College Station, Texas), separately for men and women.Cox proportional hazards analysis using the case-cohort ap-proach was used to estimate hazard ratios and 95% confi-dence intervals for the association between total CRC riskand BMI (per 5-unit increase, in sex-specific quintiles), BMIat age 20 years (per 5-unit increase, in sex-specific quintiles),BMI change (<0, 0–<4, 4–<8, or �8 units; categories werebased on a previous NLCS publication (30)), trouser/skirt size(per 2-unit size increase, in 5 sex-specific categories), andheight (per 5-cm increase, in sex-specific quintiles). To test fora linear trend across categories, we used the median anthro-pometric variable within quintiles/categories as a continuousvariable. Statistical significance was defined as P < 0.05(2-sided testing).

Multivariate models. We adjusted for potential confound-ers selected a priori from the literature and those that in-troduced more than a 10% change in hazard ratios. Selected a

1128 Hughes et al.

Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


priori were age, family history of CRC (yes/no), smoking status(never smoker, ex-smoker, or current smoker), socioeconomicstatus (educational level), total energy intake (kcal/day),alcohol intake (0, 0.1–4, 5–14, 15–29, or �30 g/day), recre-ational physical activity (<30, 30–90, or >90 minutes/day)for women, and occupational physical activity at the longest-held job (<8, 8–<12, or�12 kJ/minute) for men. No variablesintroduced more than a 10% change in the hazard ratios. BMIchange was additionally adjusted for BMI at age 20 years,and height was additionally adjusted for body weight (30).We also considered a model for women that included addi-tional adjustment for oral contraceptive use (ever, never),duration of oral contraceptive use (years), use of hormonereplacement therapy (ever, never), and duration of hormonereplacement therapy (years). This extra adjustment did notinfluence effect estimates, and results are not presented.

Interaction with physical activity. Evidence suggests thatamong persons with lower physical activity, BMI becomesa more important indicator of colon cancer risk (9) and thatunderlying population levels of physical activity can impairor enhance the ability to identify colon cancer associationswith other risk factors (31). We created a 2-way interactionbetween trouser/skirt size and physical activity. Categoriesof physical activity were based on self-reported recreationalphysical activity (women) and occupational physical activity(men), and trouser/skirt size was dichotomized on the basisof sex-specific median size. We used trouser/skirt size becauseevidence suggests that fat distribution is more important thanbody weight or BMI for CRC risk, especially in women (14).We also considered this interaction for colonic and rectaltumors separately.

For all analyses, the proportional hazards assumption wastested using scaled Schoenfeld residuals and visual inspectionof the hazard curves. Furthermore, we repeated the analysesaccording to duration of follow-up by splitting the follow-uptime at 8 years (<8 years vs. �8 years). To account for theadditional variance introduced by sampling the subcohortfrom the entire cohort, we estimated standard errors usingthe robust option.

Tests for heterogeneity. We conducted tests for hetero-geneity to evaluate differences between sublocalizationsof tumors (e.g., proximal colon vs. distal colon), using thecompeting-risks procedure in Stata. However, the standarderror for the difference in the log hazard ratios from thisprocedure assumes independence of both estimated hazardratios, which would overestimate that standard error and thusoverestimate the P value for their difference. Therefore, weestimated these P values and the associated confidenceintervals on the basis of a bootstrapping method that wasdeveloped for the case-cohort design, as described pre-viously (32). Each bootstrap analysis was based on 1,000replications.

RESULTS

Descriptive characteristics

For both men and women, a higher trouser/skirt size cor-responded with increasing BMI, and total energy intake washighest in the third quintile of BMI. Furthermore, the great-

est proportion of persons with a university-level educationwas observed in the lowest quintile of BMI, as was the greatestproportion of current smokers. In men, levels of occupationalphysical activity were relatively similar across quintiles ofBMI, whereas in women, higher levels of recreational phys-ical activity were reported by persons in the lowest quintileof BMI (Tables 1 and 2).

Associations with overall CRC in men

Age-adjusted results are not presented, since they werecomparable with multivariate results. Associations betweenbody size and CRC risk in men are presented in Table 3.With respect to body fat, BMI modeled per 5-unit increasewas associated with total CRC (hazard ratio (HR) ¼ 1.25,95% confidence interval (CI): 1.05, 1.46). When BMI wasmodeled in quintiles, associations did not reach statisticalsignificance. Stronger associations were observed with respectto trouser size; a statistically significant association was ob-served when comparing persons with the largest trouser sizeto those with the smallest size (HR ¼ 1.63, 95% CI: 1.17,2.29; P-trend ¼ 0.02).

With respect to BMI at age 20 years, BMI change, andheight, associations with overall CRC did not reach statis-tical significance.

Associations by tumor subsite in men

All indicators of body fat appeared to be most stronglyassociated with tumors of the distal colon. Increasing BMI(per 5 units) was associated with both distal (HR ¼ 1.42, 95%CI: 1.13, 1.79) and rectosigmoid (HR ¼ 1.49, 95% CI: 1.05,2.11) tumors. A statistically significant trend was observedover quintiles of BMI for these two subsites (P ¼ 0.05). BMIat age 20 years was associated only with tumors of the distalcolon (5-unit increase: HR ¼ 1.09, 95% CI: 1.00, 1.19;highest quintile vs. lowest: HR ¼ 1.47, 95% CI: 1.03, 2.08(P-trend ¼ 0.05)). With respect to a positive BMI change,there was a suggested dose-response association for tumorsof the distal colon (P-trend ¼ 0.05). The strongest associationwas observed with respect to trouser size and the distal colon(largest size vs. smallest: HR ¼ 2.56, 95% CI: 1.55, 4.24;P-trend < 0.01).

With respect to height, an inverse association was observedfor rectosigmoid tumors (5-cm increase: HR ¼ 0.77, 95% CI:0.65, 0.93; highest quintile vs. lowest: HR ¼ 0.38, 95% CI:0.17, 0.83 (P-trend ¼ 0.01)).

The strengths of the associations did not differ when datawere split according to duration of follow-up (<8 years vs.�8 years). Tests for heterogeneity between tumor subtypeswere not statistically significant for any risk factors considered.

Associations with overall CRC in women

In women, there were no statistically significant associa-tions between BMI, BMI at age 20 years, BMI change, ortrouser/skirt size and overall CRC risk (Table 4).

Height, modeled per 5-cm increase, was associated witha higher risk of total CRC (HR ¼ 1.09, 95% CI: 1.01, 1.17).There was also an association when the highest quintile was

Body Size and Risk of Colorectal Cancer 1129

Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


Table 1. Baseline Characteristics of Male Subcohort Membersa According to Quintile of Body Mass Index, Netherlands Cohort Study, 1986

Quintile of BMIb

1 (Lowest) (n 5 289) 2 (n 5 282) 3 (n 5 280) 4 (n 5 265) 5 (Highest) (n 5 249)

% Mean (SD) Median % Mean (SD) Median % Mean (SD) Median % Mean (SD) Median % Mean (SD) Median

Age at baseline, years 61.0 (4.3) 61.0 (4.1) 60.9 (4.2) 61.0 (4.2) 61.2 (4.2)

Total energy intake, kcal 2,115 (445) 2,194 (485) 2,202 (480) 2,147 (505) 2,172 (523)

Weight, kg 67.9 (6.1) 74.2 (5.6) 77.4 (5.4) 81.4 (6.3) 88.9 (8.9)

Height, cm 177.5 (6.7) 177.1 (6.6) 176.7 (6.0) 176.7 (6.5) 176.0 (6.5)

Baseline BMI 22.0 23.7 24.9 26.1 28.4

BMI at age 20 years 18.9 20.5 21.7 22.9 24.7

Trouser size 50 (2) 50 (2) 52 (2) 52 (2) 54 (2)

Occupational physicalactivity, kJ/minute

<8 63 61 62 60 60

8–<12 26 25 24 26 26

�12 11 13 14 14 14

Level of education

Primary school 18 19 23 27 29

Junior/senior high school 18 19 21 22 21

Higher vocational school 39 40 36 33 35

University 24 23 20 18 15

Smoking status

Never smoker 12 11 11 12 14

Ex-smoker 47 54 57 59 57

Current smoker 41 35 32 30 29

Alcohol intake, g/day

0 20 11 13 11 10

0.1–4 22 17 19 24 21

5–14 24 30 29 29 25

15–29 21 26 23 22 28

�30 14 16 17 15 16

Family history of colorectalcancer

No 96 95 93 96 95

Yes 4 5 7 4 5

Abbreviations: BMI, body mass index; SD, standard deviation.a Subcohort members who were not missing data for the given exposure or any of the considered baseline characteristics.b Weight (kg)/height (m)2. Range of BMI values: quintile 1, 16.1–23.0; quintile 2, 23.1–24.2; quintile 3, 24.3–25.3; quintile 4, 25.4–27.0; quintile 5, 27.1–39.6.

1130

Hughesetal.

Am

JEpidemiol.

2011;174(10):1

127–1139

at KU Leuven on November 4, 2011 http://aje.oxfordjournals.org/ Downloaded from


Table 2. Baseline Characteristics of Female Subcohort Membersa According to Quintile of Body Mass Index, Netherlands Cohort Study, 1986

Quintile of BMIb

1 (Lowest) (n 5 359) 2 (n 5 373) 3 (n 5 369) 4 (n 5 373) 5 (Highest) (n 5 358)

% Mean (SD) Median % Mean (SD) Median % Mean (SD) Median % Mean (SD) Median % Mean (SD) Median

Age at baseline, years 61.2 (4.5) 61.0 (4.3) 61.1 (4.2) 6.16 (4.2) 61.4 (4.1)

Total energy intake, kcal 1,719 (412) 1,710 (401) 1,725 (420) 1,655 (350) 1,630 (372)

Weight, kg 57.4 (5.6) 63.6 (4.6) 67.8 (4.9) 71.5 (5.6) 81.7 (8.5)

Height, cm 166.2 (6.1) 165.6 (5.9) 165.9 (5.9) 164.4 (6.2) 163.9 (5.9)

Baseline BMI 21.3 23.4 24.7 26.3 29.4

BMI at age 20 years 18.0 20.0 21.3 22.6 24.6

Trouser/skirt size 40 (2) 42 (2) 43 (2) 44 (2) 46 (2)

Recreational physicalactivity, minutes/day

<30 20 19 23 25 25

30–<60 30 32 32 33 33

60–90 25 26 23 21 21

>90 25 23 23 22 22

Level of education

Primary school 26 27 31 38 39

Junior/senior high school 17 24 26 25 28

Higher vocational school 43 38 36 29 29

University 13 12 7 8 4

Smoking status

Never smoker 49 58 54 65 63

Ex-smoker 24 18 25 21 19

Current smoker 27 24 21 14 18

Alcohol intake, g/day

0 28 31 32 33 37

0.1–4 33 35 37 39 37

5–14 22 20 18 18 15

15–29 12 11 9 8 8

�30 6 3 4 3 3

Family history of colorectalcancer

No 94 94 93 95 94

Yes 6 6 7 5 6

Abbreviations: BMI, body mass index; SD, standard deviation.a Subcohort members who were not missing data for the given exposure or any of the considered baseline characteristics.b Weight (kg)/height (m)2. Range of BMI values: quintile 1, 15.4–22.1; quintile 2, 22.2–23.8; quintile 3, 23.9–25.3; quintile 4, 25.4–27.5; quintile 5, 27.6–41.4.

BodySizeandRiskofColorectalCancer

1131

Am

JEpidemiol.

2011;174(10):1

127–1139



Table 3. Multivariate-Adjusted Hazard Ratiosa for the Association Between Anthropometric Variables and Colorectal Cancer in Men After 16.3 Years of Follow-up, by Anatomic Location,

Netherlands Cohort Study, 1986–2002

ExposurePerson-Years

at Risk

Anatomic Location

Total Colorectal Cancer Proximal Colon Distal Colon Rectosigmoid Junction Rectum

No. ofCases

HR 95% CINo. ofCases




HR 95% CI

BMIb

Per 5-unit increase 18,595 1,211 1.25 1.05, 1.46 327 1.19 0.92, 1.54 427 1.42 1.13, 1.79 125 1.49 1.05, 2.11 299 1.02 0.79, 1.32

Q1 3,772 232 1 61 1 78 1 16 1 60 1

Q2 3,931 238 0.95 0.74, 1.24 75 1.15 0.78, 1.70 78 0.92 0.64, 1.34 24 1.34 0.67, 2.61 59 0.92 0.61, 1.39

Q3 3,889 240 0.99 0.77, 1.28 58 0.89 0.59, 1.33 89 1.07 0.75, 1.53 24 1.34 0.69, 2.60 59 0.90 0.60, 1.35

Q4 3,688 247 1.05 0.81, 1.36 59 0.93 0.62, 1.40 89 1.10 0.77, 1.58 34 2.04 1.08, 3.82 60 0.93 0.62, 1.41

Q5 3,315 254 1.25 0.96, 1.62 74 1.35 0.90, 1.98 93 1.38 0.95, 1.98 27 1.62 0.85, 3.10 61 1.01 0.67, 1.51

P-trend 0.08 0.43 0.05 0.05 0.96

BMI at age 20 yearsc


Q1 3,771 238 1 69 1 84 1 21 1 50 1

Q2 4,090 240 0.92 0.72, 1.19 69 0.98 0.66, 1.44 82 0.96 0.67, 1.37 19 0.85 0.45, 1.63 62 1.18 0.77, 1.80

Q3 3,723 229 0.99 0.77, 1.28 62 0.93 0.63, 1.38 72 0.84 0.58, 1.22 26 1.24 0.68, 2.26 59 1.16 0.75, 1.79

Q4 3,622 254 1.09 0.84, 1.41 60 0.86 0.58, 1.29 89 1.03 0.72, 1.47 31 1.43 0.79, 2.60 69 1.34 0.88, 2.03

Q5 3,389 250 1.21 0.93, 1.56 67 1.05 0.72, 1.56 100 1.47 1.03, 2.08 28 1.37 0.75, 2.51 59 1.11 0.73, 1.71

P-trend 0.07 0.97 0.05 0.10 0.46

BMI changed

<0 1,646 103 0.87 0.63, 1.19 27 0.91 0.56, 1.47 36 0.87 0.56, 1.36 7 0.41 0.17, 1.02 29 0.99 0.61, 1.61

0–<4 10,298 657 1 178 1 224 1 74 1 163 1

4–<8 5,701 375 1.07 0.88, 1.30 96 0.97 0.71, 1.32 140 1.23 0.93, 1.62 37 1.01 0.63, 1.61 91 1.02 0.75, 1.40

�8 950 76 1.38 0.93, 2.05 26 1.59 0.93, 2.72 27 1.56 0.91, 2.65 7 1.29 0.51, 3.26 16 1.16 0.63, 2.15

P-trend 0.09 0.33 0.05 0.19 0.73

Height, cme

Per 5-cm increase 18,595 1,211 0.96 0.89, 1.04 327 1.01 0.90, 1.13 427 0.99 0.89, 1.11 125 0.77 0.65, 0.93 299 0.95 0.85, 1.07

Q1 5,047 337 1 94 1 96 1 32 1 72 1

Q2 2,781 167 0.85 0.65, 1.12 39 0.73 0.48, 1.12 66 1.13 0.77, 1.65 26 0.73 0.41, 1.29 62 0.84 0.56, 1.26

Q3 3,819 262 0.94 0.73, 1.20 68 0.91 0.62, 1.31 91 1.03 0.73, 1.47 34 0.90 0.51, 1.57 63 0.91 0.60, 1.37

Q4 3,480 224 0.84 0.64, 1.09 59 0.82 0.55, 1.22 84 0.96 0.65, 1.41 19 0.54 0.27, 1.06 58 0.92 0.60, 1.42

Q5 3,469 221 0.80 0.60, 1.08 67 0.90 0.59, 1.37 90 0.97 0.63, 1.48 14 0.38 0.17, 0.83 44 0.71 0.43, 1.17

P-trend 0.16 0.66 0.70 0.01 0.35

1132

Hughesetal.

Am

JEpidemiol.

2011;174(10):1

127–1139



compared with the lowest quintile (HR ¼ 1.32, 95% CI:1.03, 1.71; P-trend ¼ 0.05).

Associations by tumor subsite in women

Associations by subsite in women were not as clear asthose in men. With respect to body fat, women with thelargest trouser/skirt size had a borderline statistically signif-icant risk of proximal colon tumors compared with thosewith the smallest size (HR ¼ 1.46, 95% CI: 0.98, 2.18), butthere was no clear trend (P-trend ¼ 0.46).

There was a borderline statistically significant associationbetween height and distal colon tumors (HR ¼ 1.11, 95%CI: 0.99, 1.24) and rectal tumors (HR ¼ 1.14, 95% CI: 0.98,1.32) when data were modeled per 5-cm increase. Women inthe highest quintile also had a statistically significant risk ofdistal tumors when compared with those in the lowest quintile(HR ¼ 1.53, 95% CI: 1.03, 2.27; P-trend ¼ 0.05). Thisassociation was borderline statistically significant for rectaltumors (HR ¼ 1.49, 95% CI: 0.98, 2.27; P-trend ¼ 0.06).

When data were split at 8 years of follow-up, it appearedthat associations for all anthropometric variables were at-tenuated over time (data not shown). Tests for heterogeneitybetween tumor subtypes were not statistically significant forany risk factors considered.

Interaction between trouser/skirt size andphysical activity

Table 5 shows hazard ratios for the interaction betweentrouser size and occupational physical activity in men. Com-pared with the reference category of men with a small trousersize and a high level of physical activity, there were nostatistically significant associations for total CRC, but menwith a small trouser size and a low level of physical activitywere at increased risk of distal colon tumors (HR ¼ 1.63,95% CI: 1.03, 2.56). A borderline statistically significantassociation was also observed for men with a large trousersize and all levels of physical activity with respect to tumorsof the distal colon.

Compared with the reference category of women with asmall trouser/skirt size and a high level of physical activity(Table 6), the greatest risk for each subtype considered appearedto be for persons with the highest trouser/skirt size and thelowest level of physical activity (for total CRC, HR ¼ 1.83,95% CI: 1.28, 2.63; for proximal colon tumors, HR ¼ 1.70,95% CI: 1.08, 2.67; for distal colon tumors, HR ¼ 1.95, 95%CI: 1.21, 3.17; for rectal tumors, HR ¼ 2.56, 95% CI: 1.36,4.79). The interaction between trouser/skirt size and physicalactivity was statistically significant only for the proximalcolon (P < 0.05).

DISCUSSION

We observed that in men, body fat, as indicated by BMI,BMI at age 20 years, BMI change, and trouser size, appearedto be more associated with distal tumors than with tumors atother anatomic subsites; however, associations with respectto height were less clear. Body size appeared to be unrelatedto CRC outcomes in women, except for height, which was

Trousersize(D

utchsize

in1986)f

Per2-sizeincrease

18,595

1,211

1.12

1.05,1.19

327

1.09

0.99,1.20

427

1.16

1.04,1.29

125

1.07

0.93,1.22

299

1.06

0.96,1.17

Q1

2,760

128

143

132

115

135

1

Q2

4,563

280

1.31

0.98,1.74

70

0.96

0.62,1.48

95

1.78

1.14,2.78

25

1.04

0.53,2.04

83

1.42

0.90,2.23

Q3

6,185

421

1.46

1.11,1.91

116

1.21

0.81,1.80

156

2.14

1.40,3.27

48

1.51

0.82,2.79

87

1.09

0.70,1.63

Q4

3,269

239

1.56

1.15,2.10

59

1.16

0.74,1.82

87

2.23

1.40,3.54

26

1.51

0.76,3.01

63

1.50

0.94,2.39

Q5

1,819

143

1.63

1.17,2.29

39

1.32

0.81,2.15

57

2.56

1.55,4.24

11

1.10

0.49,2.49

31

1.33

0.77,1.29

P-trend

0.02

0.14

<0.01

0.28

0.34

Abbreviatio

ns:BMI,bodymassindex;CI,confidenceinterval;HR,hazard

ratio

;Q,quintile.

aHazard

ratio

swere

adjustedforage,totalenergyintake(kcal/d

ay),occupatio

nalphysicalactivity

(<8,8

–<12,o

r�12kJ/m

inute),levelofe

duca

tion(primary

school,somehighschool,high

vocatio

nalschool,oruniversity),family

history

ofcolorectalcancer(yes/no),alcoholconsumptio

n(0,0

.1–4,5

–14,o

r15–29g/day),a

ndsmoking(neversmoker,ex-smoker,orcurrentsmoker).

bWeight(kg)/height(m

)2.RangeofBMIvalues:

Q1,16.1–23.0;Q2,23.1–24.2;Q3,24.3–25.3;Q4,25.4–27.0;Q5,27.1–39.6.

cRangeofBMIvaluesatage20years:Q1,11.8–19.8;Q2,19.9–21.0;Q3,21.1–22.1;Q4,22.2–23.6;Q5,23.7–31.8.

dIn

themodelforBMIchange,hazard

ratio

swere

additionally

adjustedforBMIatage20years.Bodyweightchangeassociatedwith

categoriesofBMIchange:�29to

0kg;0.1–15kg;

11–30kg;23–57kg.

eIn

themodelforheight,hazard

ratio

swere

additionally

adjustedforbodyweight(kg).Rangeofheightvalues:

Q1,158–172cm;Q2,173–175cm;Q3,176–178cm;Q4,179–182cm;

Q5,183–200cm.

fRangeoftrousersizes:Q1,28–48;Q2,50;Q3,52;Q4,54;Q5,56–68.


Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


Table 4. Multivariate-Adjusted Hazard Ratiosa for the Association Between Anthropometric Variables and Colorectal Cancer Endpoints in Women After 16.3 Years of Follow-up, by Anatomic

Location, Netherlands Cohort Study, 1986–2002

ExposurePerson-Years

at Risk

Anatomic Location

Total Colorectal Cancer Proximal Colon Distal Colon Rectosigmoid Junction Rectum

No. ofCases





HR 95% CI

BMIb


Q1 5,238 228 1 94 1 70 1 16 1 42 1

Q2 5,623 211 0.88 0.69, 1.13 87 0.88 0.63, 1.23 62 0.85 0.58, 1.24 17 1.08 0.53, 2.21 41 1.00 0.63, 1.58

Q3 5,499 223 0.94 0.73, 1.20 94 0.99 0.71, 1.40 58 0.81 0.55, 1.19 26 1.86 0.94, 3.69 38 0.87 0.54, 1.40

Q4 5,606 222 0.91 0.71, 1.16 95 0.98 0.70, 1.37 64 0.85 0.58, 1.23 13 0.85 0.40, 1.84 40 0.93 0.58, 1.49

Q5 5,229 222 0.97 0.76, 1.24 89 0.91 0.65, 1.28 73 1.04 0.72, 1.50 15 1.06 0.50, 2.26 44 1.07 0.67, 1.60

P-trend 0.90 0.84 0.84 0.93 0.90

BMI at age 20 yearsc


Q1 5,628 213 1 90 1 60 1 19 1 38 1

Q2 5,247 227 1.21 0.95, 1.56 100 1.28 0.90, 1.76 67 1.22 0.83, 1.79 18 1.09 0.56, 1.13 43 1.26 0.78, 2.01

Q3 5,212 236 1.25 0.98, 1.61 96 1.15 0.82, 1.60 74 1.36 0.95, 1.99 18 0.97 0.50, 1.89 46 1.34 0.84, 2.13

Q4 5,576 212 1.08 0.84, 1.38 87 0.98 0.70, 1.38 65 1.21 0.82, 1.79 15 0.85 0.42, 1.72 35 0.98 0.60, 1.61

Q5 5,532 218 1.12 0.87, 1.43 86 1.14 0.81, 1.62 61 1.09 0.74, 1.62 17 1.07 0.54, 2.12 43 1.29 0.80, 2.07

P-trend 0.68 0.97 0.68 0.89 0.62

BMI changed

<0 3,268 134 0.94 0.73, 1.22 51 0.86 0.60, 1.23 44 1.08 0.72, 1.60 12 1.26 0.63, 2.50 19 0.67 0.39,1.14

0–<4 12,177 507 1 212 1 147 1 37 1 97 1

4–<8 8,787 343 0.89 0.74, 1.06 144 0.86 0.67, 1.10 98 0.89 0.67, 1.20 26 0.97 0.56, 1.68 69 0.96 0.68, 1.35

�8 2,963 122 0.95 0.73, 1.24 52 0.98 0.68, 1.40 38 1.00 0.67, 1.50 12 1.40 0.69, 2.82 20 0.83 0.49, 1.39

P-trend 0.55 0.86 0.58 0.83 0.78

Height, cme

Per 5 cm 27,195 1,106 1.09 1.01, 1.17 459 1.04 0.95, 1.15 327 1.11 0.99, 1.24 87 1.02 0.82, 1.25 205 1.14 0.98, 1.32

Q1 6,353 250 1 104 1 68 1 24 1 49 1

Q2 4,932 186 0.96 0.75, 1.23 89 1.11 0.80, 1.54 62 1.17 0.80, 1.72 10 0.53 0.25, 1.12 20 0.52 0.30, 0.89

Q3 5,599 209 0.97 0.77, 1.24 103 1.15 0.83, 1.58 49 0.84 0.56, 1.25 13 0.61 0.30, 1.24 40 0.93 0.60, 1.45

Q4 5,708 231 1.03 0.81, 1.31 78 0.83 0.59, 1.18 76 1.28 0.88, 1.86 23 1.02 0.55, 1.86 45 1.04 0.68, 1.59

Q5 4,603 230 1.32 1.03, 1.71 85 1.19 0.84, 1.70 72 1.53 1.03, 2.27 17 0.90 0.44, 1.84 51 1.49 0.98, 2.27

P-trend 0.05 0.67 0.05 0.78 0.06

1134

Hughesetal.

Am

JEpidemiol.

2011;174(10):1

127–1139



related positively to cancers of the distal colon and rectum.In general, our observations, especially in men, highlight theimportance of examining CRC risks according to tumorsubsite, because stronger associations were observed for in-dividual subsites than with overall risk.

Strengths of this study include the prospective cohort de-sign and the large population, in combination with sufficientfollow-up time to allow precise analysis of CRC risk atdifferent subsites. The NLCS has almost complete ascertain-ment of CRC cases, and follow-up of the subcohort is almost100%. Although measures of body size in this study wereobtained by self-report, there are many examples in the lit-erature showing that this method is a reasonably valid andreliable tool for assessing body weight and height in cohortstudies (33–36). A weakness of this study was the smallnumber of rectosigmoid and rectal tumors and the subse-quently limited statistical power.

Investigators in several case-control studies have consid-ered the association between BMI and site-specific tumorsof the colon in men, and they have generally reported a stron-ger risk for tumors of the distal colon than for tumors of theproximal colon (6–10). This is supported by the results of arecent prospective cohort study carried out by Laake et al. (11).In our study, we considered several indicators of body fat inaddition to BMI, including BMI at age 20 years, BMI change,and trouser/skirt size as a proxy for waist circumference,and we consistently observed an elevated risk of distal colontumors as compared with tumors at other subsites. Althoughwe did not observe statistically significant heterogeneitybetween proximal and distal colon tumors as Laake et al. (11)did, the bootstrapping method we used is quite conservative.The weaker associations that we observed with respect toassociations between body fat and the rectum align withprevious research (1), but we also observed a positive associ-ation between BMI and tumors of the rectosigmoid junction.To our knowledge, no other prospective study has consideredthe association between anthropometric measures and therectosigmoid junction. This association is plausible, becausethe rectosigmoid junction and distal colon are anatomicallylinked. Our findings should be replicated in other prospectivestudies, but they support the hypothesis that the etiology ofCRC tumors differs between anatomic subsites of the co-lorectal tract in men.

We did not observe clear associations between BMI, BMIat age 20 years, or BMI change and CRC risk in women, butour results suggest that women with a high trouser/skirt sizeare at increased risk of proximal colon tumors, although thisassociation did not reach statistical significance. In contrast,positive associations between BMI and distal colon tumorshave been reported in prospective cohort studies of Swedish(12) and Norwegian (11) women. In an American cohort, bothBMI and waist circumference were associated with proximaland distal tumors (13). The proportion of obese women in theNLCS is small, and perhaps this prevented us from detectingassociations of similar strength. It has also been suggested thatthe impact of a given risk factor along the length of the largebowel may differ according to the prevalence of other envi-ronmental factors and, thus, according to sex and country (2).

Alternatively, it is plausible that a metabolic profile re-flecting a combination of risk factors has a greater influence

Trouser/skirtsize(D

utch

sizein

1986)f

Per2-sizeincrease

27,195

1,106

1.03

0.98,1.09

459

1.06

0.98,1.14

327

1.02

0.94,1.11

87

1.03

0.90,1.18

205

1.04

0.94,1.16

Q1

5,317

210

180

166

115

136

1

Q2

6,861

284

1.02

0.80,1.29

126

1.20

0.86,1.67

80

0.89

0.62,1.28

20

1.07

0.54,2.13

52

1.08

0.68,1.71

Q3

7,586

283

0.90

0.71,1.14

114

0.95

0.68,1.33

83

0.83

0.58,1.19

29

1.38

0.73,2.60

52

0.96

0.61,1.53

Q4

4,761

197

0.98

0.76,1.27

76

0.97

0.67,1.39

57

0.92

0.62,1.36

19

1.39

0.68,2.87

44

1.28

0.79,2.08

Q5

2,670

132

1.17

0.87,1.58

63

1.46

0.98,2.18

41

1.15

0.74,1.80

40.58

0.19,1.78

21

1.07

0.59,1.93

P-trend

0.62

0.46

0.72

0.90

0.56

Abbreviatio

ns:BMI,bodymassindex;CI,confidenceinterval;HR,hazard

ratio

;Q,quintile.

aHazard

ratio

swere

adjustedforage,totale

nergyintake(kcal/d

ay),recreatio

nalp

hysicala

ctivity

(<30,30–<60,60–90,or>90minutes/day),levelofeducatio

n(primary

school,somehigh

school,highvoca

tionalschool,oruniversity),family

history

ofcolorectalcancer(yes/no),alcoholconsumptio

n(0,0.1–4,5–14,or15–29g/day),andsmoking(neversmoker,ex-smoker,or

currentsmoker).

bWeight(kg)/height(m

)2.RangeofBMIvalues:Q1,15.4–22.1;Q2,22.2–23.8;Q3,23.9–25.3;Q4,25.4–27.5;Q5,27.6–41.4.

cRangeofBMIvaluesatage20years:Q1,11.3–19.0;Q2,19.1–20.6;Q3,20.7–21.8;Q4,21.9–23.3;Q5,23.4–46.9.

dIn

themodelforBMIchange,h

aza

rdratio

swere

additionally

adjustedforBMIa

tage20ye

ars.B

odyweightchangeassociatedwith

categoriesofB

MIchange:�

50to

0kg

;0.1–12kg

;9–25kg

;

18–62kg

.eIn

themodelforheight,hazard

ratio

swere

additionally

adjustedforbodyweight(kg).Rangeofheightvalues:

Q1,143–160cm;Q2,161–164cm;Q3,165–167cm;Q4,168–170cm;Q5,

171–186cm.

fRangeoftrouser/skirtsizes:

Q1,36–40;Q2,42;Q3,44;Q4,46;Q5,48–56.


Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


on cancer risk than do the individual effects of body size andobesity (37, 38). Because evidence suggests that fat distri-bution is more important than body weight or BMI for CRCrisk, especially in women (14), we investigated the interactionwith physical activity using trouser/skirt size. Our observa-tions for women were intriguing. As noted above, we did not

observe statistically significant associations with respect to trou-ser/skirt size in our general analysis. However, when we consid-ered the interaction between trouser/skirt size and physicalactivity, it appeared that women with a large trouser/skirt sizeand the lowest level of physical activity were at the greatestrisk of CRC compared with women with a small trouser/skirt

Table 5. Multivariate-Adjusted Hazard Ratiosa for the Interaction Between Occupational Physical Activity and Trouser Size in the Risk of

Colorectal Cancer Among Men, Netherlands Cohort Study, 1986–2002

Trouser Sizeb

Physical Activity

Low Intermediate High

No. ofCases

PY atRisk


PY atRisk


PY atRisk

HR 95% CI

Total colorectal cancer

�52 632 10,139 1.20 0.90, 1.60 249 4,392 1.03 0.76, 1.39 123 2,293 1

>52 263 3,821 1.15 0.83, 1.61 123 1,490 1.38 0.95, 2.02 75 940 1.34 0.87, 2.06

Proximal colon

�52 169 10,139 1.03 0.67, 1.58 83 4,392 1.10 0.70, 1.72 37 2,293 1

>52 81 3,821 1.14 0.70, 1.84 29 1,490 1.07 0.61, 1.89 15 940 0.84 0.42, 1.67

Distal colon

�52 233 10,139 1.63 1.03, 2.56 66 4,392 1.10 0.70, 1.72 30 2,293 1

>52 103 3,821 1.54 0.95, 2.54 41 1,490 1.72 0.98, 3.03 26 940 1.76 0.94, 3.30

Rectum

�52 140 10,139 0.94 0.61, 1.45 68 4,392 0.92 0.59, 1.45 40 2,293 1

>52 51 3,821 0.89 0.52, 1.53 37 1,490 1.51 0.86, 2.66 22 940 1.41 0.75, 2.67

Abbreviations: CI, confidence interval; HR, hazard ratio; PY, person-years.a Hazard ratios were adjusted for trouser size, physical activity, total energy intake, family history of colorectal cancer, socioeconomic status,

alcohol consumption, and smoking status.b Dutch size in 1986.

Table 6. Multivariate-Adjusted Hazard Ratiosa for the Interaction Between Nonoccupational Physical Activity and Trouser/Skirt Size in the Risk of

Colorectal Cancer Among Women, Netherlands Cohort Study, 1986–2002

Trouser/Skirt Sizeb

Physical Activity

Low Intermediate High

No. ofCases

PY atRisk


PY atRisk


PY atRisk

HR 95% CI

Total colorectal cancer

�44 159 4,471 1.45 1.08, 1.94 357 12,026 1.25 0.97, 1.60 122 5,054 1

>44 97 2,283 1.83 1.28, 2.63 116 4,447 1.14 0.82, 1.59 51 1,617 1.44 0.95, 2.18

Proximal colon*

�44 89 4,471 1.60 1.10, 2.31 203 12,026 1.40 1.02, 1.93 63 5,054 1

>44 52 2,283 1.70 1.08, 2.67 66 4,447 1.16 0.76, 1.77 33 1,617 1.69 1.03, 2.79

Distal colon

�44 63 4,471 1.28 0.86, 1.91 141 12,026 1.08 0.77, 1.52 54 5,054 1

>44 42 2,283 1.95 1.21, 3.17 50 4,447 1.18 0.74, 1.84 17 1,617 1.12 0.61, 2.03

Rectum

�44 38 4,471 1.67 0.98, 2.85 86 12,026 1.43 0.89, 2.29 25 5,054 1

>44 27 2,283 2.56 1.36, 4.79 35 4,447 1.70 0.93, 3.10 12 1,617 1.66 0.77, 3.54

Abbreviations: CI, confidence interval; HR, hazard ratio; PY, person-years.

* P for interaction < 0.05.a Hazard ratios were adjusted for trouser/skirt size, physical activity, total energy intake, family history of colorectal cancer, socioeconomic

status, alcohol consumption, and smoking status.b Dutch size in 1986.

1136 Hughes et al.

Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


size and a high level of physical activity. This risk was highestfor rectal tumors; however, the interaction was statisticallysignificant only for tumors of the proximal colon. Thesefindings require confirmation in other prospective studies, butthey support case-control data (9, 39). They also highlight thatunderlying population levels of physical activity can impairor enhance the ability to identify CRC associations withother risk factors (31), such as body fat.

Height is a reflection of earlier life exposures, such aschildhood energy intake (40). In our study, height was as-sociated with increased risk of CRC in women, which is inaccordance with previous research (1, 13, 14, 41, 42). In arecent report, Oxentenko et al. (13) suggested that risk isgreater for tumors of the proximal colon, whereas our datasuggest that risk is greater for tumors of the distal colon.However, in the former study, Oxentenko et al. did not reportconducting tests for heterogeneity, and in our study, such testsdid not reach statistical significance. We did not observe anyassociation with respect to height and CRC risk in men, andwe actually observed an inverse association for tumors of therectosigmoid junction. We cannot explain this observation,and we suggest that it may be a chance finding. We have shownpreviously in the NLCS population that energy restrictionduring childhood is inversely related to later CRC risk (43);therefore, while we have not observed strong associationsbetween height and CRC in the present study, it is plausiblethat early-life nutritional factors may influence differentcarcinogenic pathways than later-life exposure to increasedbody fat (13).

We can only speculate as to why body size and body fat inparticular might differentially influence the etiology of colo-rectal tumors at different subsites. Obesity is associated with achronic state of low-grade inflammation and thus increasedcirculatory levels of inflammatory markers like C-reactiveprotein, interleukin-6, and tumor necrosis factor a (44). Thesein turn may induce insulin resistance and hyperinsulinemia.It is hypothesized that such conditions influence CRC risk(45–47); however, it remains unclear how they directly orindirectly influence specific tumor subsites. There are somedistinct differences between the proximal colon and the distalcolon, both anatomically and genetically (3–5). The distalcolon is associated more with physical/chemical stimuli andassociated less with water absorption and electrolyte trans-port than the proximal colon (3). It has been shown thattumors characterized by chromosomal instability are moreassociated with tumors of the distal colon, whereas epigeneticchanges like methylation are more associated with tumors ofthe proximal colon (48). Interestingly, recent studies havesuggested that overweight and obesity do not appear to dif-ferentially influence CRC risk via epigenetic mechanismsand microsatellite instability (20, 22, 49). Based on the ob-servation that body fat appears to be associated with a higherrisk of distal colon tumors, a plausible hypothesis is thatobesity and its associated process may influence risk via thechromosomal instability pathway. However, to our knowl-edge, this has not been investigated in any population-basedstudies. It is clear that the association between metabolic andhormonal risk factors and CRC is complex, and more researchin this area is needed to elucidate clear mechanisms and howthese mechanisms might differ with respect to tumor subsite.

In this Dutch population, body fat—especially abdominalfat, as indicated by trouser size—was associated with tumorsof the distal colon in men. In women, the influence of bodyfat on CRC risk may differ according to level of physicalactivity. Height was associated with CRC risk only inwomen. More research is needed to fully elucidate howbody size may influence subsite-specific pathways of CRCcarcinogenesis.

ACKNOWLEDGMENTS

Author affiliations: Department of Epidemiology, GROWSchool for Oncology and Developmental Biology, MaastrichtUniversity, Maastricht, the Netherlands (Laura A. E. Hughes,Colinda C. J. M. Simons, Piet A. van den Brandt, Matty P.Weijenberg); Department of Pathology, GROW School forOncology and Developmental Biology, Maastricht University,Maastricht, the Netherlands (Manon van Engeland); and De-partment of Prevention and Health, TNO Quality of Life,Leiden, the Netherlands (R. Alexandra Goldbohm).

This study was supported by a grant from the World CancerResearch Fund (grant 2007/54) to Dr. Matty P. Weijenberg.

The authors thank the staffs of the municipal cancer reg-istries (IKA, IKL, IKMN, IKN, IKO, IKR, IKST, IKW, IKZ,and VIKC) and the Netherlands nationwide pathologyregistry (PALGA). They also thank S. van de Crommertand J. Nelissen for their assistance with data entry and datamanagement; A. Volovics and A. Kester for statistical advice;Dr. L. Schouten, C. de Zwart, M. Moll, W. van Dijk, M. Jansen,and A. Pisters for data management assistance; and H. vanMontfort, T. van Moergastel, L. van den Bosch, and R. Schmeitzfor programming assistance.

Conflict of interest: none declared.

REFERENCES

1. World Cancer Research Fund/American Institute for CancerResearch. Food, Nutrition, Physical Activity, and the Preventionof Cancer: A Global Perspective. Washington, DC: AmericanInstitute for Cancer Research; 2007.

2. Morois S, Mesrine S, Josset M, et al. Anthropometric factorsin adulthood and risk of colorectal adenomas: the French E3N-EPIC prospective cohort. Am J Epidemiol. 2010;172(10):1166–1180.

3. Li FY, Lai MD. Colorectal cancer, one entity or three. J ZhejiangUniv Sci B. 2009;10(3):219–229.

4. Breivik J, Lothe RA, Meling GI, et al. Different geneticpathways to proximal and distal colorectal cancer influencedby sex-related factors. Int J Cancer. 1997;74(6):664–669.

5. Lindblom A. Different mechanisms in the tumorigenesis ofproximal and distal colon cancers. Curr Opin Oncol. 2001;13(1):63–69.

6. Caan BJ, Coates AO, Slattery ML, et al. Body size and the riskof colon cancer in a large case-control study. Int J Obes RelatMetab Disord. 1998;22(2):178–184.

7. Gerhardsson de Verdier M, Hagman U, Steineck G, et al. Diet,body mass and colorectal cancer: a case-referent study inStockholm. Int J Cancer. 1990;46(5):832–838.


Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


8. Russo A, Franceschi S, La Vecchia C, et al. Body size andcolorectal-cancer risk. Int J Cancer. 1998;78(2):161–165.

9. Slattery ML, Potter J, Caan B, et al. Energy balance and coloncancer—beyond physical activity.CancerRes.1997;57(1):75–80.

10. West DW, Slattery ML, Robison LM, et al. Dietary intake andcolon cancer: sex- and anatomic site-specific associations. AmJ Epidemiol. 1989;130(5):883–894.

11. Laake I, Thune I, Selmer R, et al. A prospective study of bodymass index, weight change, and risk of cancer in the proximaland distal colon. Cancer Epidemiol Biomarkers Prev. 2010;19(6):1511–1522.

12. Terry P, Giovannucci E, Bergkvist L, et al. Body weight andcolorectal cancer risk in a cohort of Swedish women: relationvaries by age and cancer site. Br J Cancer. 2001;85(3):346–349.

13. Oxentenko AS, Bardia A, Vierkant RA, et al. Body size andincident colorectal cancer: a prospective study of olderwomen. Cancer Prev Res (Phila). 2010;3(12):1608–1620.

14. Pischon T, Lahmann PH, Boeing H, et al. Body size and risk ofcolon and rectal cancer in the European Prospective Investi-gation Into Cancer and Nutrition (EPIC). J Natl Cancer Inst.2006;98(13):920–931.

15. Pischon T, Nothlings U, Boeing H. Obesity and cancer. ProcNutr Soc. 2008;67(2):128–145.

16. Derks S, Postma C, Carvalho B, et al. Integrated analysis ofchromosomal, microsatellite and epigenetic instability in co-lorectal cancer identifies specific associations between pro-moter methylation of pivotal tumour suppressor and DNArepair genes and specific chromosomal alterations. Carcino-genesis. 2008;29(2):434–439.

17. Barault L, Charon-Barra C, Jooste V, et al. Hypermethylatorphenotype in sporadic colon cancer: study on a population-based series of 582 cases. Cancer Res. 2008;68(20):8541–8546.

18. Samowitz WS, Albertsen H, Herrick J, et al. Evaluation of alarge, population-based sample supports a CpG island meth-ylator phenotype in colon cancer. Gastroenterology. 2005;129(3):837–845.

19. Weisenberger DJ, Siegmund KD, Campan M, et al. CpG islandmethylator phenotype underlies sporadic microsatellite insta-bility and is tightly associated with BRAF mutation in colo-rectal cancer. Nat Genet. 2006;38(7):787–793.

20. Campbell PT, Jacobs ET, Ulrich CM, et al. Case-control studyof overweight, obesity, and colorectal cancer risk, overall andby tumor microsatellite instability status. Colon Cancer Fam-ily Registry. J Natl Cancer Inst. 2010;102(6):391–400.

21. Hughes LA, Simons CC, van den Brandt PA, et al. Body size,physical activity and risk of colorectal cancer with or withoutthe CpG Island Methylator Phenotype (CIMP). PloS One.2011;6(4):e18571. (doi:10.1371/journal.pone.0018571).

22. Slattery ML, Curtin K, Sweeney C, et al. Diet and lifestylefactor associations with CpG island methylator phenotype andBRAF mutations in colon cancer. Int J Cancer. 2007;120(3):656–663.

23. van den Brandt PA, Goldbohm RA, van ‘t Veer P, et al. A large-scale prospective cohort study on diet and cancer in theNetherlands. J Clin Epidemiol. 1990;43(3):285–295.

24. Goldbohm RA, van den Brandt PA, Brants HA, et al. Valida-tion of a dietary questionnaire used in a large-scale prospectivecohort study on diet and cancer. Eur J Clin Nutr. 1994;48(4):253–265.

25. Goldbohm RA, van ‘t Veer P, van den Brandt PA, et al. Re-producibility of a food frequency questionnaire and stability ofdietary habits determined from five annually repeated mea-surements. Eur J Clin Nutr. 1995;49(6):420–429.

26. Van den Brandt PA, Schouten LJ, Goldbohm RA, et al. De-velopment of a record linkage protocol for use in the Dutch

Cancer Registry for Epidemiological Research. Int J Epidemiol.1990;19(3):553–558.

27. Goldbohm RA, van den Brandt PA, Dorant E. Estimation ofthe coverage of Dutch municipalities by cancer registriesand PALGA based on hospital discharge data. Tijdschr SocGezondheidsz. 1994;72:80–84.

28. Hughes LA, Schouten LJ, Goldbohm RA, et al. Self-reportedclothing size as a proxy measure for body size. Epidemiology.2009;20(5):673–676.

29. Hettinger TH, Mueller BH, Gebhard H. Ermittlung des Ar-beitsenergieumsatzes bei Dynamisch Muskulaerer Arbeit.Dortmund, Germany: Schriftenreihe der Bundesarbeit furArbeitsschutz; 1989.

30. Pylypchuk RD, Schouten LJ, Goldbohm RA, et al. Body massindex, height, and risk of lymphatic malignancies: a prospec-tive cohort study. Am J Epidemiol. 2009;170(3):297–307.

31. Slattery ML, Potter JD. Physical activity and colon cancer:confounding or interaction? Med Sci Sports Exerc. 2002;34(6):913–919.

32. de Vogel S, Bongaerts BW, Wouters KA, et al. Associations ofdietary methyl donor intake with MLH1 promoter hyperme-thylation and related molecular phenotypes in sporadic colo-rectal cancer. Carcinogenesis. 2008;29(9):1765–1773.

33. Bolton-Smith C, Woodward M, Tunstall-Pedoe H, et al. Ac-curacy of the estimated prevalence of obesity from self re-ported height and weight in an adult Scottish population.J Epidemiol Community Health. 2000;54(2):143–148.

34. Nyholm M, Gullberg B, Merlo J, et al. The validity of obesitybased on self-reported weight and height: implications for pop-ulation studies. Obesity (Silver Spring). 2007;15(1):197–208.

35. Spencer EA, Appleby PN, Davey GK, et al. Validity of self-reported height and weight in 4808 EPIC-Oxford participants.Public Health Nutr. 2002;5(4):561–565.

36. Wada K, Tamakoshi K, Tsunekawa T, et al. Validity of self-reported height and weight in a Japanese workplace popula-tion. Int J Obes (Lond). 2005;29(9):1093–1099.

37. McKeown-Eyssen G. Epidemiology of colorectal cancer re-visited: are serum triglycerides and/or plasma glucose asso-ciated with risk? Cancer Epidemiol Biomarkers Prev. 1994;3(8):687–695.

38. Silvera SA, Jain M, Howe GR, et al. Energy balance and breastcancer risk: a prospective cohort study. Breast Cancer Res Treat.2006;97(1):97–106.

39. Slattery ML, Caan BJ, Benson J, et al. Energy balance and rectalcancer: an evaluation of energy intake, energy expenditure, andbody mass index. Nutr Cancer. 2003;46(2):166–171.

40. Okasha M, Gunnell D, Holly J, et al. Childhood growth andadult cancer. Best Pract Res Clin Endocrinol Metab. 2002;16(2):225–241.

41. Engeland A, Tretli S, Austad G, et al. Height and body massindex in relation to colorectal and gallbladder cancer in twomillion Norwegian men and women. Cancer Causes Control.2005;16(8):987–996.

42. Wei EK, Giovannucci E, Wu K, et al. Comparison of risk factorsfor colon and rectal cancer. Int J Cancer. 2004;108(3):433–442.

43. Hughes LA, van den Brandt PA, Goldbohm RA, et al. Child-hood and adolescent energy restriction and subsequent colo-rectal cancer risk: results from the Netherlands Cohort Study.Int J Epidemiol. 2010;39(5):1333–1344.

44. Monteiro R, Azevedo I. Chronic inflammation in obesity andthe metabolic syndrome. Mediators Inflamm.2010;2010:289645. (doi:10.1155/2010/289645).

45. Elwing JE, Gao F, Davidson NO, et al. Type 2 diabetes mel-litus: the impact on colorectal adenoma risk in women. Am JGastroenterol. 2006;101(8):1866–1871.

1138 Hughes et al.

Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


46. Giovannucci E, Michaud D. The role of obesity and relatedmetabolic disturbances in cancers of the colon, prostate, andpancreas. Gastroenterology. 2007;132(6):2208–2225.

47. Hu FB, Manson JE, Liu S, et al. Prospective study of adultonset diabetes mellitus (type 2) and risk of colorectal cancer inwomen. J Natl Cancer Inst. 1999;91(6):542–547.

48. Jass JR. Classification of colorectal cancer based on correla-tion of clinical, morphological and molecular features. Histo-pathology. 2007;50(1):113–130.

49. Slattery ML, Wolff RK, Curtin K, et al. Colon tumor mutationsand epigenetic changes associated with genetic polymorphism:insight into disease pathways. Mutat Res. 2009;660(1-2):12–21.


Am J Epidemiol. 2011;174(10):1127–1139

at KU

Leuven on N

ovember 4, 2011


ownloaded from


body weight and risk of colon cancer

Documents